Patient Speci fi c Computer Modelling for Automated Sizing of Fenestrated Stent Grafts
Lucie Deryckea,b,*, Jean Sénémaudb, David Perrinc, Stephane Avrila, Pascal Desgrangesb, Jean-Noel Albertinid, Frederic Cochennecb, Stephan Haulone
aMines Saint-Etienne, Univ Lyon, Univ Jean Monnet, INSERM, U 1059 Sainbiose, Centre CIS, Fe42023 Saint-Etienne, France
bDepartment of Vascular Surgery, Henri Mondor Hospital, University of Paris XII, Créteil, France
cPrediSurge, 3, Saint-Etienne, France
dDepartment of Cardio-Vascular Surgery, Centre Hospitalier Régional Universitaire de Saint-Etienne, Saint-Priez-en-Jarez, France
eDepartment of Aortic and Vascular Surgery, Marie Lannelongue Hospital, Le Plessis-Robinson, INSERM UMR_S 999, Université Paris Sud, France
WHAT THIS PAPER ADDS
Fenestrated stent grafting has become a standard endovascular approach to treat complex abdominal aortic aneurysms. Currently, accurate device planning requires significant operator experience and a long manufacturing delay. This step is known to be crucial for surgical success. Computational modelling could prove helpful in providing more tailored and straightforward care in patients, especially for planning. This study ex- plores the potential of simulation for fenestrated stent graft sizing.
Objectives: The aim was to validate a computational patient specific model of ZenithÒ fenestrated device deployment in abdominal aortic aneurysms to predict fenestration positions.
Methods: This was a retrospective analysis of the accuracy of numerical simulation for fenestrated stent graft sizing. Finite element computational simulation was performed in 51 consecutive patients that underwent successful endovascular repair with ZenithÒfenestrated stent grafts in two vascular surgery units with a high volume of aortic procedures. Longitudinal and rotational clock positions of fenestrations were measured on the simulated models. These measurements were compared with those obtained by (i) an independent observer on the post-operative computed tomography (CT) scan and (ii) by the stent graft manufacturer planning team on the pre-operative CT scan. (iii) Pre- and post-operative positions were also compared.
Longitudinal distance and clock face discrepancies >3 mm and 15, respectively, were considered significant.
Reproducibility was assessed using BlandeAltman and linear regression analysis.
Results: A total of 195 target arteries were analysed. Both BlandeAltman and linear regression showed good reproducibility between the three measurement techniques performed. The median absolute difference between the simulation and post-operative CT scan was 1.0 1.1 mm for longitudinal distance measurements and 6.9 6.1 for clock positions. The median absolute difference between the planning centre and post-operative CT scan was 0.8 0.8 mm for longitudinal distance measurements and 5.15.0 for clock positions. Finally, the median absolute difference between the simulation and the planning centre was 0.96 0.97 mm for longitudinal distance measurements and 4.8 3.6 for clock positions.
Conclusions: The numerical model of deployed fenestrated stent grafts is accurate for planning position of fenestrations. It has been validated in 51 patients, for whom fenestration locations were similar to the sizing performed by physicians and the planning centre.
Keywords:Computational analysis, Fenestrated endovascular aneurysm repair, Numerical simulation personalised medicine Article history: Received 14 April 2019, Accepted 16 October 2019, Available online XXX
Ó2019 European Society for Vascular Surgery. Published by Elsevier B.V. All rights reserved.
INTRODUCTION
In most high volume European aortic centres, stent graft (SG) implantation is considered the first line treatment for thoracic and abdominal aortic aneurysms (AAAs) in patients with suitable anatomies.1,2 Fenestrated stent grafting (FEVAR) is a validated endovascular approach to treat com- plex AAA with unfavourable anatomies (short infrarenal neck, suprarenal or type IV thoraco-abdominal aneurysms).3e7
* Corresponding author. 51 Avenue du Maréchal de Lattre de Tassigny, 94010 Créteil, France.
E-mail address:deryckelucie@gmail.com(Lucie Derycke).
1078-5884/Ó2019 European Society for Vascular Surgery. Published by Elsevier B.V. All rights reserved.
https://doi.org/10.1016/j.ejvs.2019.10.009
FEVAR requires custom made devices specifically tailored to each patient’s anatomy. As a result, the delay for planning and manufacturing currently ranges from six to eight weeks, which may preclude its use in patients presenting with large aneurysms. Accurate device planning requires the use of dedicated three dimensional imaging software combined with a high resolution pre-operative computed tomography (CT) scan to perform all the necessary measurements. In patients with arterial tortuosity or important angulation, it is difficult to predict SG behaviour in the aorta and to modify the semi-automated arterial segmentation and centreline reconstructions accordingly.
Finite element analysis (FEA) is a numerical method used to overcome biomechanical problems using displacement, strain, and stress analysis.8FEA can be used to predict the deployment of SGs in aortic aneurysms.9e13 Recent litera- ture has highlighted the advances in computational analysis applied to endovascular repair. Based on analyses focused on the mechanical behaviour of SGs,14models of EVAR and suprarenal devices deployed in virtual models have been established;15,16these patient specific models have proved reliable.9e13
The aim of this study was to assess two different steps of a patient specific finite element model of ZenithÒ fenes- trated device deployment in complex AAAs to predict fenestration positions. The primary hypothesis was that simulation would provide accurate positioning of fenestra- tions compared with standard planning techniques and post-operative CT scan measurements.
MATERIALS AND METHODS
Study population
The study was approved by the Institutional Review Board of the French Society of Thoracic and Cardiovascular Sur- gery (Société Française de Chirurgie Thoracique et Cardio VasculairedSFCTCV).
The study population consisted of 51 consecutive pa- tients who underwent implantation of a ZenithÒ fenes- trated AAA endovascular graft (Cook Medical, Bloomington, IN) to treat juxtarenal, pararenal, or thoraco-abdominal aortic aneurysms between January 2016 and April 2018 at Henri Mondor Hospital and Marie Lannelongue Hospital.
Exclusion criteria were non-availability of pre-operative CT scan, poor pre-operative CT scan quality precluding modelling analysis (poor quality of the contrast agent in- jection or large slice thickness (>3 mm)), post-dissection aneurysms, and devices combining branch(es) and fenestration(s).
Sizing process
The sizing of the SG was performed by the manufacturer (Cook MedicalÒ, London, UK).
Simulation strategy
High resolution pre-operative arterial phase CT scans were used to generate patient specific three dimensional (3D)
models of the aorta, from the descending thoracic aorta to the common iliac arteries, including the target arteries (coeliac trunk, superior mesenteric, and renal arteries) using the VMTK (Vascular Modeling Toolkit, www.vmtk.org) li- brary. This open source software semi-automatically seg- ments DICOM (Digital Imaging and Communications in Medicine) datasets from the CT scan and creates 3D vascular models (aortic segmentation step).17 As calcifica- tions of the aortic wall and intraluminal thrombus were not included in the model, the aortic wall and the target vessels were considered as homogeneous surfaces with a constant thickness of 1.5 mm and 1 mm respectively. The custom made SGs were reproduced digitally on the basis of the graft plan data. Deployment of the personalised SG in its corresponding model of the aorta was performed retro- spectively using the commercially available Abaqus/Explicit v6.14 finite element solver (Dassault Systèmes, Paris, France). Simulations were achieved using proprietary algo- rithms (Planop, PrediSurge, Saint-Etienne, France). Compu- tational technology was based on finite element analysis and assessed the deformations induced by the deviceehost interaction resulting in a prediction of SG behaviour and arterial displacement. Briefly, the aortic surface was virtu- ally deformed until a cylindrical shape of the aorta was obtained. This step of numerical aortic deformation was called “morphing”. Then, it was possible to deploy a stent graft inside the cylindrical aorta, and to reverse the defor- mation of the aortic surface while maintaining the stent graft in place (Fig. 1). The validation of this model was previously reported by the group.9,10
The aortic wall was assigned an orthotropic elastic behaviour. Mechanical properties of stents and fabric were obtained either from literature or from in house mechanical testing of samples obtained from manufacturers.14For the stents, a linear elastic material behaviour was used, repro- ducing the nitinol behaviour in its austenitic phase.14 The polyester fabric was modelled as an orthotropic elastic material.14
Simulation analysis
For each visceral artery incorporated into the graft, two data parameters were measured by a blinded investigator:
the longitudinal distance between the ostial centres along the centreline, measured in millimetres (mm), and the rotational clock angle between each visceral artery, measured in degrees (). These two parameters encoded the position of the visceral arteries relative to one another.
Automated fenestration positions were calculated by the software from the“morphing”step.
The different steps of the simulation methodology are illustrated inFig. 1.
Post-operative images analysis
One trained and independent observer performed the post- operative imaging analysis. The position of the fenestrations on the post-operative CT scan represents the “ideal” SG configuration because any sizing mismatch is compensated
by the stent graft’s longitudinal (stents are not connected to each other) and rotational (10%e20% oversizing)flexibility.
The measures of relative angle and distance between each visceral artery were extracted using a dedicated imaging workstation (TeraRecon Inc., Santa Rosa, CA, USA) and the modality previously described for pre-operative sizing.18
Statistical analysis
All statistical analyses were performed using R statistical software version 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria). Unlike the real deployment process, the simulated SG was not deployed inside the aorta based on the position of one of the target arteries but with an automatic longitudinal position and an overall rotation of zero degrees. A systematic measurement bias could have been induced by this method. In order to provide relevant comparisons between the implanted FEVAR and simulation processes, a constant shift was applied to all the fenestra- tions for each case to correct potential differences induced by these two parameters. However, the relative distances and angles between the fenestrations for each case were preserved, as illustrated inFig. 2.
The measures were compared with respect to the abso- lute longitudinal and rotational differences.
Agreement between the different measures was assessed by plotting the difference between each method relative to
each other with the limits of agreement (1.96 standard deviations around the mean difference) as described by Bland and Altman. Quantitative variables were also ana- lysed by linear regression test, used to calculate the bias, estimated by the mean difference and the standard devia- tion of the difference.
Longitudinal distance and clock face discrepancies >
3 mm and>15, respectively, were considered significant, according to the interobserver variability of pre-operative sizing reported in the literature.18e20
RESULTS
Fifty-one FEVAR deployments were simulated with a total of 180 fenestrations and 15 scallops (195 target arteries).
Details are reported inTable 1. Examples of simulation re- sults are presented inFig. 3. Fifteen patients were excluded for non-availability of pre-operative CT scan (3/15), poor pre-operative CT scan quality (5/15), post-dissection aneu- rysms (4/15), and devices combining branch(es) and fen- estration(s) (3/15).
In current practice, numerical simulations required approximately eight hours of computational analysis, including two hours of manual work. The automated fenestration positions extracted from the“morphing”step were obtained in approximately half an hour.
Aortic model
Stent graft
model Deployment simulaion Assessment
Morphing
Morphing results
Simulation results
Pre-operative sizing
Post-operative sizing Simulation deployment
1/2 H 8 H
(2H for user)
A B C
Figure 1.Simulation methodology. (A) Aortic and stent graft modelling (B)“Morphing”step and deployment simulation. (C) Fenestrations positioning on the cylindrical aortic shape and on the deformed graft are compared with standard planning techniques and post-operative computed tomography scan measurements.
The mean slice thickness of the pre-operative CT scans was 1.20.45 mm (0.5e2.0). Similar results were obtained when comparing infra- or millimetric CT scans and supra- millimetric CT scans.
Quantitative assessment
The numerical results are summarised inTable 2.
Both BlandeAltman and linear regression showed good reproducibility for longitudinal and circumferential position between the three methods (Fig. 4).
The median absolute difference between the simulation and post-operative CT scan was 1.01.1 mm for longitu- dinal distance measurements and 6.96.1 for clock po- sitions. The worst case was a longitudinal distance difference of 6.0 mm and an angle difference of 44.3. Ninety five per cent of the longitudinal deviances were shorter than 3 mm (43/51 cases) and 96% of the rotational deviances were under 15 (44/51 cases).
The median absolute difference between the planning centre and post-operative sizing was 0.8 0.8 mm for longitudinal distance measurements and 5.1 5.0 for clock positions. The worst case was a longitudinal distance difference of 4.0 mm and an angle difference of 37.1.
Ninety seven per cent of the longitudinal deviances were shorter than 3 mm (46/51 cases) and 98% of the rotational deviances were under 15 (48/51 cases).
The median absolute difference between the simulation and the planning centre was 0.96 0.97 mm for longitu- dinal distance measurements and 4.8 3.6 for clock po- sitions. The worst case was a longitudinal distance difference of 5.0 mm and an angle difference of 21.8. Ninety eight per cent of the longitudinal deviances were shorter than 3 mm (47/51 cases) and 99% of the rotational deviances were less than 15 (49/51 cases).
A subgroup analysis focusing on the number of fenes- trations (group 1: four fenestrations or scallop (39/51) and group 2: three fenestrations or scallop (11/51)) showed similar results. The median absolute difference between the simulation and the post-operative sizing was approximately 1 mm (1.01.2 mm and 1.01.2 mm, respectively) for longitudinal distance measurements and 7.5 (7.0 6.3 and 7.95.2, respectively) for clock positions. The median absolute difference between the simulation and the plan- ning centre was approximately 1 mm (0.9 1.2 mm and 1.20.9 mm, respectively) for longitudinal distance mea- surements and 5 (4.83.6 and 4.83.8, respectively) for clock positions.
Qualitative assessment
Two cases of proximal suboptimal apposition were observed and one case presented with an excess graft oversizing (Fig. 5A,B). An inhomogeneous deployment of the SG was observed in six cases; it was also depicted on the post-operative CT scan. On the example, the posterior stent struts were tightened whereas the anterior struts were spread apart. This stent configuration was similar on the simulation and the post-operative CT scan (Fig. 5C). Two cases presented a SG shifted towards the aortic axis because of a significant misalignment of the aneurysm sac and the aortic axis (Fig. 6). In six cases, pleated fabric was
Table 1. Distribution of fenestrations and scallops of the simulated stent grafts
nof fenestrations/
scallops
Cases (n)
Fenestrations (n)
Scallops (n)
5/0 1 5 0
4/0 28 112 0
3/0 7 21 0
4/1 1 4 1
3/1 10 30 10
2/1 4 8 4
Total 51 180 15
–180° –90° 0° 90° 180°
Longitudinale position
Circumferential position
Pre- and post-operative positions Validation Zone
Simulation positions Shifted simulation positions Constant shift
Figure 2.Illustration of constant shift applied on simulation results. The relative distances and angles between the fenestrations for each case are preserved.
observed at the level of the visceral arteries and a similar aspect was found on the corresponding post-operative CT scan (Fig. 5D,E).
Automated sizing process
The numerical results are summarised inTable 2.
The median absolute difference between automated and post-operative sizing was 3.0 0.3 mm for longitudinal distance measurements and 11.09.3for clock positions.
The worst case was a longitudinal distance difference of 16.7 mm and an angle difference of 56.0. Ninety three per
cent of the longitudinal deviances were shorter than 3 mm (39/51 cases) and 91% of the rotational deviances were under 15 (38/51 cases).
The median absolute difference between automated sizing and the planning centre was 1.21.7 mm for lon- gitudinal distance measurements and 7.46.5 for clock positions. The worst case of each was a longitudinal dis- tance difference of 15.5 mm and an angle difference of 44.9. Ninety three per cent of the longitudinal deviances were under 3 mm (40/51 cases) and 97% of the rotational deviances were under 15 (46/51 cases).
A B C
Figure 3.Illustration of the three different stent graft modelling and simulation results. (A) stent graft with proximal bare stent. (B,C) Stent graft without bare stent with two different geometries.
DISCUSSION
Treatment of complex AAA by FEVAR has been proven feasible, effective and safe3e7 and has the potential to greatly reduce the mortality and morbidity risks, especially
in the subgroup of patients at high risk of open repair.
However, the use of fenestrated devices implies proper analysis of the arterial geometry to design a custom made SG. Physicians have to be trained to use dedicated sizing
Table 2.Comparison results of longitudinal and clock positions obtained by the two steps of the simulation model and the pre- and post-operative sizing; and percentage of longitudinal and clock position discrepancies below the significance limits of 3 mm and 15
Longitudinal positionemm Circumferential positione
Post-operative sizing Pre-operative sizing Post-operative sizing Pre-operative sizing Median±SD
(range)
n£3 mm (%)
Median±SD (range)
n£3 mm (%)
Median±SD (range)
n£15 (%)
Median±SD (range)
n£15 (%)
Simulation 1.01.1
(e5.9 to 6.0)
95 0.960.97
(e4.6 to 5.0)
98 6.96.1
(e44.3 to 25.1)
96 4.83.6
(e21.8 to 19.3) 99 Pre-operative sizing 0.80.8
(e4.0 to 4.0)
97 5.15.0
(e37.1 to 18.4) 98 Automated positions 3.00.3
(e9.5 to 16.7)
93 1.21.7
(e15.5 to 9.5)
93 11.09.3
(e56.0 to 38.0)
91 6.56.1
(e44.9 to 34.0) 93 SD¼standard deviation.
–140
–140 –120 –100 –80 –60 –40 –20 –120
–100 –80 –60 –40 –20
–100 –50 0 50 100 150
–100 –50 0 50 100 150 Linear Regression (y=0.99x–0.84)
Simulation (mm) Longitudinal position
Simulation (degree) Circumferential position Post-operative sizing (mm) Longitudinal positionPost-operative sizing (degree) Circumferential position
Linear Regression (y=0.89x+1.05) y=x line
regression line
y=x line regression line
A
D
–140
–140 –120 –100 –80 –60 –40 –20 –120
–100 –80 –60 –40 –20
–100 –50 0 50 100 150
–100 –50 0 50 100 150 Linear Regression (y=0.99x–0.41)
Pre-operative sizing (mm) Longitudinal position
Pre-operative sizing (degree) Circumferential position Post-operative sizing (degree) Circumferential positionPost-operative sizing (mm) Longitudinal position
Linear Regression (y=0.93x–0.41) y=x line
regression line
y=x line regression line
B
E
–140
–140 –120 –100 –80 –60 –40 –20 –120
–100 –80 –60 –40 –20
–100 –50 0 50 100 150
–100 –50 0 50 100 150 Linear Regression (y=0.99x–0.41)
Simulation (mm) Longitudinal position
Simulation (degree) Circumferential position Post-operative sizing (degree) Circumferential positionPost-operative sizing (mm) Longitudinal position
Linear Regression (y=0.95x–0.78) y=x line
regression line
y=x line regression line
C
F
Figure 4.Longitudinal and clock position of target vessels shown with the simulation and the pre- and post-operative measurements. The comparison is performed by linear regression analysis. (AeC) Comparison of longitudinal positions extracted from simulation and post- operative sizing (A), from pre- and post-operative sizing (B), and from simulation and pre-operative sizing (C). (DeF) Comparison of circumferential positions extracted from simulation and post-operative sizing (D), from pre- and post-operative sizing (E), and from simu- lation and pre-operative sizing (F).
software and to be familiar with 3D imaging and its limi- tations.21,22An automated, reproducible and precise sizing procedure could facilitate the clinical process and free up physician time.
The model proved successful in providing fenestration locations similarly to the current sizing performed by the planning centre, confirmed by the comparison with the post-operative analysis. The process has been successfully evaluated in 51 patients.
Accurate positioning of the fenestrations is a determinant of procedural success and long term patency. Numerical simulation of SG deployment offers other potential advan- tages. The behaviour of the deployed SG can be visualised and assessed, including adequate application of the fabric, proximal apposition, excessive oversizing or kinking, as seen in Fig. 5. These results indicate the clinical relevance of numerical simulation of FEVAR to improve outcome by helping the physician to select appropriate graft diameters and fenestration positions that bestfits the individual pa- tient. In unfavourable aortic anatomies, a deployment simulation could be performed by the planning centre to accept or reject cases. To allow relevant comparison, an accurate digital reproduction of patient specific SGs was performed based on the graft plan data. Standardised dig- ital SG models are currently being constructed to provide a quick and easy to use interface.
In less than 15% of cases longitudinal and rotational differences greater than 3 mm or 15were recorded. There
is no consensus regarding maximum tolerance level be- tween fenestrations and target vessels positions. Some experienced endovascular specialists suggested that a rotational deviance of 15between the fenestration and its target vessel was acceptable, as it was unlikely to lead to significant complications.23The results confirm the previous studies focused on interobserver variability when sizing FEVAR, which reported good agreement but also some critical discrepancies between observers. Oshin et al. re- ported an analysis of 25 FEVAR sizings performed by two experienced operators; they observed deviances of more than 3 mm in longitudinal position in 18% of cases, and of more than 15 in 12% of cases on clock position.20 Banno et al.18 reported a comparison of 268 FEVAR sizings be- tween two experienced surgeons and the manufacturer.
They observed more than 22.5 angle discrepancy in 9.8%
of cases and more than 5 mm length discrepancy in 16.4%
of cases. Finally, Malkawi et al.19 reported results of a comparison of 19 FEVAR sizings among four experienced observers. The overall interobserver measurement error for distance was 5.3 mm (95% confidence interval (CI) 4.4e6.2) and for target vessel orientation 12.6(95% CI 10.8e14.4).
Maurel et al.24 reported a comparison of FEVAR pre- and post-operative sizing of renal artery distance and clock po- sition and showed that an accommodation to sizing error up to 15 in clock position may be considered acceptable, without adverse consequences on patency. The current re- sults compare favourably with these data.
A B C
D E
Figure 5.Qualitative assessment examples. (A) Oversizing excess; (B) defect of proximal appo- sition; (C) case of inhomogeneous stent graft deployment: simulation (in red) is compared with a similar view of the implanted stents extracted from the post-operative computed tomography (CT) scan (in grey); left lateral and posterior views. (D,E) Anterior view of two cases with a pleated fabric on simulations results (in red) and comparison with similar view of the actual stents extracted from the post-operative CT scan (in grey).
Furthermore, discrepancies could be partly explained by the adjustment of the fenestrations designed by the manufacturer tofit with the 15%e20% graft oversizing, the presence of reducing ties, and fenestration shift to avoid having a stent strut crossing the lumen of a fenestration.
These rotational adjustments explain why longitudinal measurements showed less variability than circumferential measurements. The technique of clock position measure- ment, with its 7.5 accuracy, compared with the 1 mm ac- curacy of longitudinal measurements, also explains this difference.
A thorough examination of the cases has identified cau- ses of discrepancy. An inhomogeneous deployment of the SG was observed in six cases associated with rotational position misalignment, which was also found on the post- operative CT scan (Fig. 5). In misalignment of the aneu- rysmal sac and aortic axis, the SG was shifted away from the aortic axis, especially in cases with prior EVAR. InFig. 6, one of these cases is presented: the SG, which is constrained above and below the aneurysm, is shifted forward in the
aortic lumen at the level of the aneurysm. Different clock positions are obtained depending on whether the aortic lumen axis or the SG axis is used as a reference. In six cases with a longitudinal position misalignment, pleated fabric was observed at the level of the visceral arteries and a similar aspect was found on the corresponding post- operative CT scan, as shown in Fig. 5. Finally, discrep- ancies were shown in two cases with a large distance be- tween the SG and the visceral artery ostium and three cases with large visceral artery diameter (> 10 mm). In such anatomies, accuracy is less crucial to achieve technical success.
The delay in producing a fenestrated device is related to both planning and manufacturing. The current eight week delay is inappropriate for urgent patients. Nowadays nu- merical simulations require approximately eight hours of computational analysis, including two hours of manual work, which is similar to the time spent for planning by a trained operator. To further reduce the delay and provide an accurate and fast tool to physicians, computing capabilities
A B
C
Figure 6.Challenging clock position analysis. (A) Right lateral view; red arrow, the stent graft has shifted forwards in the aortic lumen; black arrow, the stent graft is constrained by the previously implanted endovascular abdominal aortic aneurysm repair. Transverse views of (B) superior mesenteric and right renal arteries clock positions with the clock aligned with the aortic lumen centreline and (C) with the clock aligned with the stent graft lumen centreline.
improve at a very rapid pace. In the near future, it is likely that the time required to perform these simulations will fall below an hour.
The first step of the deployment simulation provides visceral artery positions within half an hour. However, the aortic segmentation was inaccurate in some cases with arterial calcification, previous EVAR with struts in front of visceral arteries and ostial stenosis. For such cases, checking the aortic segmentation is still required. A previous study highlighted the need for“human input”in automated sizing for EVAR.25Measurement accuracy appeared to be better in the current study, probably due to the improved quality of the CT scans.
This study has several limitations. The study focused on fenestrated devices and non-dissected aortas. Branched device modelling is currently feasible and will be assessed in specific studies. Improvements are also in progress to extend the model to dissected aortas. Moreover, the study focused on the Zenith Cook Medical device because it is the most commonly implanted fenestrated endograft in both centres, and because the position of the fenestration is critical on this device, which is fully supported with stents.
Simulation of other fenestrated devices, which require different planning methods, will be performed in a separate study. The simulations could benefit from more sophisti- cated modelling, including thrombus, and aortic wall calci- fication and environment. Current validation of simulation models remains a major issue. Simulation results were compared with pre- and post-operative sizing, which have their own limitations, and CT images analysis and numerical models of SG deployment were performed retrospectively.
Numerical simulation of SG deployment in complex AAAs was feasible in 51 patients and accurately determined pa- tient specific fenestration positions. When compared with post-operative sizing, the current process of sizing achieved by the planning centre showed less variability than the simulation model and automated positions extracted directly from the “morphing” step still required manual corrections. It is currently being improved with new computational algorithms. This model is a reliable tool for FEVAR planning and has the potential to help physicians select the device design that best fits the patient’s aortic anatomy. Evaluation of the software in other conditions, such as unsuccessful cases, high arterial tortuosity cases and physician modified stent grafts, should be addressed in further studies. Moreover, prospective work should be done in order to support these promising early results.
CONFLICT OF INTEREST
J.N. Albertini, D. Perrin, and S. Avril are cofounders of the company PrediSurge SAS. S. Haulon is consultant for Cook Medical. F. Cochennec is proctor for Cook Medical. The other authors have no conflict of interest.
FUNDING None.
REFERENCES
1 Wanhainen A, Verzini F, Van Herzeele I, Allaire E, Bown M, Cohnert T, et al. Editor’s choice - European Society for Vascular Surgery (ESVS) 2019 clinical practice guidelines on the manage- ment of abdominal aorto-iliac artery aneurysms. Eur J Vasc Endovasc Surg2019;57:8e93.
2 Riambau V, Böckler D, Brunkwall J, Cao P, Chiesa R, Coppi G, et al. Editor’s choice - management of descending thoracic aorta diseases: clinical practice guidelines of the European Society for Vascular Surgery (ESVS).Eur J Vasc Endovasc Surg2017;53:4e52.
3 Mastracci TM, Eagleton MJ, Kuramochi Y, Bathurst S, Wolski K.
Twelve-year results of fenestrated endografts for juxtarenal and group IV thoracoabdominal aneurysms.J Vasc Surg2015;61:355e 64.
4 Marzelle J, Presles E, Becquemin JP. WINDOWS trial participants.
Results and factors affecting early outcome of fenestrated and/or branched stent grafts for aortic aneurysms: a multicenter pro- spective study.Ann Surg2015;261:197e206.
5 Verhoeven ELG, Vourliotakis G, Bos WTGJ, Tielliu IFJ, Zeebregts CJ, Prins TR, et al. M. Fenestrated stent grafting for short-necked and juxtarenal abdominal aortic aneurysm: an 8- year single-centre experience.Eur J Vasc Endovasc Surg2010;39:
529e36.
6 Oderich GS, Greenberg RK, Farber M, Lyden S, Sanchez L, Fairman R, et al. Results of the United States multicenter pro- spective study evaluating the Zenith fenestrated endovascular graft for treatment of juxtarenal abdominal aortic aneurysms.
J Vasc Surg2014;60:1420e8. e1e5.
7 British Society for Endovascular Therapy and the Global Collab- orators on Advanced Stent-Graft Techniques for Aneurysm Repair (GLOBALSTAR) Registry. Early results of fenestrated endovas- cular repair of juxtarenal aortic aneurysms in the United Kingdom.Circulation2012;125:2707e15.
8 Leach JR, Mofrad MRK, Saloner D. Computational models of vascular mechanics [Internet]. In: Computational modeling in biomechanics. Dordrecht: Springer Netherlands; 2010. p. 99e 170.
9 Perrin D, Badel P, Orgeas L, Geindreau C, du Roscoat SR, Albertini J-N, et al. Patient-specific simulation of endovascular repair surgery with tortuous aneurysms requiringflexible stent- grafts.J Mech Behav Biomed Mater2016;63:86e99.
10 Perrin D, Badel P, Orgéas L, Geindreau C, Dumenil A, Albertini J- N, et al. Patient-specific numerical simulation of stent-graft deployment: validation on three clinical cases. J Biomech 2015;48:1868e75.
11 Auricchio F, Conti M, Marconi S, Reali A, Tolenaar JL, Trimarchi S. Patient-specific aortic endografting simulation: from diagnosis to prediction.Comput Biol Med2013;43:386e94.
12 Romarowski RM, Conti M, Morganti S, Grassi V, Marrocco- Trischitta MM, Trimarchi S, et al. Computational simulation of TEVAR in the ascending aorta for optimal endograft selection: a patient-specific case study.Comput Biol Med2018;103:140e7.
13 Derycke L, Perrin D, Cochennec F, Albertini J-N, Avril S. Predic- tive numerical simulations of double branch stent-graft deploy- ment in an aortic arch aneurysm.Ann Biomed Eng2019.
14 Demanget N, Avril S, Badel P, Orgéas L, Geindreau C, Albertini J- N, et al. Computational comparison of the bending behavior of aortic stent-grafts.J Mech Behav Biomed Mater2012;5:272e82.
15 Perrin D, Demanget N, Badel P, Avril S, Orgéas L, Geindreau C, et al. Deployment of stent grafts in curved aneurysmal arteries:
toward a predictive numerical tool.Int J Numer Methods Biomed Eng2015;31:e02698.
16 De Bock S, Iannaccone F, De Beule M, Vermassen F, Segers P, Verhegghe B. What if you stretch the IFU? A mechanical insight into stent graft Instructions For Use in angulated proximal aneu- rysm necks.Med Eng Phys2014;36:1567e76.
17 Sethian JA. A fast marching level set method for monotonically advancing fronts.Proc Natl Acad Sci U S A1996;93:1591e5.
18 Banno H, Kobeiter H, Brossier J, Marzelle J, Presles E, Becquemin J-P. Inter-observer variability in sizing fenestrated and/or branched aortic stent-grafts.Eur J Vasc Endovasc Surg Off J Eur Soc Vasc Surg2014;47:45e52.
19 Malkawi AH, Resch TA, Bown MJ, Manning BJ, Poloniecki JD, Nordon IM, et al. Sizing fenestrated aortic stent-grafts.Eur J Vasc Endovasc Surg2011;41:311e6.
20 Oshin OA, England A, McWilliams RG, Brennan JA, Fisher RK, Vallabhaneni SR. Intra- and interobserver variability of target vessel measurement for fenestrated endovascular aneurysm repair.J Endovasc Ther2010;17:402e7.
21 Moore R, Hinojosa CA, O’Neill S, Mastracci TM, Cinà CS. Fenes- trated endovascular grafts for juxtarenal aortic aneurysms: a step by step technical approach.Catheter Cardiovasc Interv 2007;69:
554e71.
22 Greenberg RK, West K, Pfaff K, Foster J, Skender D, Haulon S, et al. Beyond the aortic bifurcation: branched endovascular grafts for thoracoabdominal and aortoiliac aneurysms. J Vasc Surg 2006;43:879e86. discussion 886-887.
23 Nordon IM, Hinchliffe RJ, Manning B, Ivancev K, Holt PJ, Loftus IM, et al. Toward an“off-the-shelf”fenestrated endograft for management of short-necked abdominal aortic aneurysms: an analysis of current graft morphological diversity.J Endovasc Ther 2010;17:78e85.
24 Maurel B, Lounes Y, Amako M, Fabre D, Hertault A, Sobocinski J, et al. Changes in renal anatomy after fenestrated endovascular aneurysm repair.Eur J Vasc Endovasc Surg2017;53:95e102.
25 Wyss TR, Dick F, England A, Brown LC, Rodway AD, Greenhalgh RM. Three-dimensional imaging core laboratory of the endovascular aneurysm repair trials: validation of methodol- ogy.Eur J Vasc Endovasc Surg2009;38:724e31.
